Chat Completions

POST https://api.shannon-ai.com/v1/chat/completions

Eng kichik so'rov — model id va bitta foydalanuvchi xabari.

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://api.shannon-ai.com/v1",
)

response = client.chat.completions.create(
    model="shannon-3",
    messages=[{"role": "user", "content": "Say hello in one sentence."}],
)

print(response.choices[0].message.content)

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "YOUR_API_KEY",
  baseURL: "https://api.shannon-ai.com/v1",
});

const response = await client.chat.completions.create({
  model: "shannon-3",
  messages: [{ role: "user", content: "Say hello in one sentence." }],
});

console.log(response.choices[0].message.content);

curl https://api.shannon-ai.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "shannon-3",
    "messages": [{"role": "user", "content": "Say hello in one sentence."}]
  }'

Javob — bitta JSON obyekti:

{
  "id": "chatcmpl-5f0c1e7a9b3d4c62a8e1f07d2b46c9a3",
  "object": "chat.completion",
  "created": 1791625200,
  "model": "shannon-3",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Hello, it is good to meet you.",
        "reasoning_content": "The user wants a greeting in one sentence. Keep it short and friendly."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 1184,
    "completion_tokens": 46,
    "total_tokens": 1230
  }
}

Sarlavhalar

So'rov sarlavhalari

Javob sarlavhalari

So'rov maydonlari

Faqat messages majburiy. Qo'llaydi ustuni maydon javobni o'zgartiradigan modellarni nomlaydi. Xostingdagi open-weight modellar — model ro'yxatidagi o'n ikkita id; Shannon 3 oilasi — shannon-3, shannon-3-pro, shannon-3.1 va shannon-3.1-pro. Modellar va narxlar

Boshqa OpenAI maydonlari, masalan n, user, stream_options, parallel_tool_calls, presence_penalty, frequency_penalty, logit_bias, logprobs, metadata, store va prompt_cache_key, mavjud mijoz kodi o'zgarishsiz ishlashi uchun qabul qilinadi. Ular javobni o'zgartirmaydi: doim bitta choice bor va oqim doim usage bilan tugaydi.

Noto'g'ri JSON turidagi maydon, masalan "max_tokens": "100", 422 qaytaradi. messagessiz so'rov ham shunday.

Tool'lar, tuzilmali chiqish, mulohaza va veb-qidiruvning har birining o'z sahifasi bor: Funksiya chaqirish, Tuzilgan chiqishlar, Mulohaza darajasi, O‘rnatilgan veb qidiruv.

Parametrlar bilan so'rov

Bu so'rov system xabari, sampling maydonlari va mulohaza darajasini belgilaydi. U xostingdagi open-weight modeldan foydalanadi, u hammasini qo'llaydi.

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://api.shannon-ai.com/v1",
)

response = client.chat.completions.create(
    model="DeepSeek-V4-Flash-0731-W4A16-AUTOROUND-REAP",
    messages=[
        {"role": "system", "content": "You are a physics teacher. Answer in two sentences."},
        {"role": "user", "content": "Why is the sky blue?"},
    ],
    max_tokens=512,
    temperature=0.3,
    top_p=0.9,
    seed=7,
    stop=["\n\n"],
    reasoning_effort="low",
)

message = response.choices[0].message
print(message.reasoning_content)  # the reasoning
print(message.content)            # the answer
print(response.usage)

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "YOUR_API_KEY",
  baseURL: "https://api.shannon-ai.com/v1",
});

const response = await client.chat.completions.create({
  model: "DeepSeek-V4-Flash-0731-W4A16-AUTOROUND-REAP",
  messages: [
    { role: "system", content: "You are a physics teacher. Answer in two sentences." },
    { role: "user", content: "Why is the sky blue?" },
  ],
  max_tokens: 512,
  temperature: 0.3,
  top_p: 0.9,
  seed: 7,
  stop: ["\n\n"],
  reasoning_effort: "low",
});

const message = response.choices[0].message;
console.log(message.reasoning_content); // the reasoning
console.log(message.content);           // the answer
console.log(response.usage);

curl https://api.shannon-ai.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "DeepSeek-V4-Flash-0731-W4A16-AUTOROUND-REAP",
    "messages": [
      {"role": "system", "content": "You are a physics teacher. Answer in two sentences."},
      {"role": "user", "content": "Why is the sky blue?"}
    ],
    "max_tokens": 512,
    "temperature": 0.3,
    "top_p": 0.9,
    "seed": 7,
    "stop": ["\n\n"],
    "reasoning_effort": "low"
  }'

Javob yuqoridagi bilan bir xil shaklda. Uning usage'i xostingdagi open-weight modellarda ikkita tafsilot qo'shadi: keshdan o'qilgan prompt tokenlari va mulohazaga sarflangan tokenlar.

{
  "usage": {
    "prompt_tokens": 31,
    "completion_tokens": 62,
    "total_tokens": 93,
    "prompt_tokens_details": {
      "cached_tokens": 0
    },
    "completion_tokens_details": {
      "reasoning_tokens": 21
    }
  }
}

Chiqish uzunligi

max_tokens ikki ish qiladi. Birinchidan, so'rov boshlanganda balansingizdan ajratib qo'yiladigan tokenlar soni shu. Javob to'liq bo'lgach, bu miqdor so'rov ishlatgan tokenlar bilan almashtiriladi. Agar max_tokens balansingizda qolgan miqdordan katta bo'lsa, javobning o'zi sig'sa ham so'rov 429 Quota exceeded qaytaradi. Kamroq ajratish uchun pastroq max_tokens yuboring.

shannon-coder-1 bu endpointda boshqacha hisoblanadi: har bir so'rov rejangizdagi Shannon Coder chaqiruvlaridan biri bo'ladi va uning uchun token ajratib qo'yilmaydi. Limitlar va balans

Ikkinchidan, u bu modellarda javob uzunligini cheklaydi:

max_tokens yoki max_completion_tokens bo'lmasa, qiymat 4,096. shannon-coder-1 da u 65,536.

Xabarlar

Har bir xabar — role va contentli obyekt. content — satr, xabar matndan ko'proq narsani olib kelsa esa qismlar massivi.

Shannon 3 oilasi id'sida bajarilishi shart bo'lgan ko'rsatmalarni user xabariga qo'ying.

Shannon darajalarida foydalanuvchi matni ham, tools ham bo'lmagan so'rov 400 No user message provided qaytaradi.

Kontent qismlari

O'lchamlar, limitlar va shakllarning to'liq ro'yxati alohida sahifada. Rasmlar va fayllar

Javob obyekti

Tugash sabablari

Oqimsiz javob stop yoki tool_callsni ko'rsatadi.

Usage

Xostingdagi open-weight modellarda prompt_tokens — xabarlaringiz va tool ta'riflari modelning o'z tokenizatori bilan sanalgani, qo'shimcha ravishda rasmlar tokenlari. Token sanash endpointlari yuborishdan oldin xuddi shu sonni qaytaradi. Tokenlarni sanash

Shannon darajalarida prompt_tokens model javob yozish uchun o'qigan hamma narsani sanaydi, shuning uchun u faqat xabarlaringiz matnidan kattaroq bo'ladi.

Streaming

stream true bo'lganda javob chat.completion.chunk hodisalari sifatida keladi va data: [DONE] bilan tugaydi. Undan oldingi oxirgi chunk finish_reason va usageni olib keladi; stream_options kerak emas. Chunk shakllari, keep-alive satrlari va oqim ichidagi xatolar alohida sahifada. Oqim

Xatolar

Xato — error a'zosi bor JSON obyekti. Tekshiruvlar shu tartibda bajariladi: API kalit, so'rov tanasi, model id, so'ng balans. Jadvalda bu endpoint eng ko'p qaytaradigan xatolar keltirilgan. To'liq ro'yxat, nimani qayta urinish kerakligi bilan, alohida sahifada. Xatolarni boshqarish

{
  "error": {
    "type": "invalid_request_error",
    "message": "unknown model: no-such-model"
  }
}